arXiv subjects

David R. Cheriton

Publications and source records attributed to David R. Cheriton.

3 recordsLinked to original sources

arXiv2005

Enforcing Bulk Mail Classification

Spam costs US corporations upwards of $8.9 billion a year, and comprises as much as 40% of all email received. Solutions exist to reduce the amount of spam seen by end users, but cannot withstand sophisticated attacks. Worse yet, many will occasionally misclassify and silently drop legitimate email. Spammers take advantage of the near-zero cost of sending email to flood the network, knowing that success even a tiny fraction of the time means a profit. End users, however, have proven unwilling to pay money to send email to friends and family. We show that it is feasible to extend the existing mail system to reduce the amount of unwanted email, without misclassifying email, and without charging well-behaved users. We require that bulk email senders accurately classify each email message they send as an advertisement with an area of interest or else be charged a small negative incentive per message delivered. Recipients are able to filter out email outside their scope of interest, while senders are able to focus their sendings to the appropriate audience.

cs.NI↗

arXiv2005

A Scalable Stream-Oriented Framework for Cluster Applications

This paper presents a stream-oriented architecture for structuring cluster applications. Clusters that run applications based on this architecture can scale to tenths of thousands of nodes with significantly less performance loss or reliability problems. Our architecture exploits the stream nature of the data flow and reduces congestion through load balancing, hides latency behind data pushes and transparently handles node failures. In our ongoing work, we are developing an implementation for this architecture and we are able to run simple data mining applications on a cluster simulator.

cs.DC↗

arXiv2003

Active Internet Traffic Filtering: Real-time Response to Denial of Service Attacks

Denial of Service (DoS) attacks are one of the most challenging threats to Internet security. An attacker typically compromises a large number of vulnerable hosts and uses them to flood the victim's site with malicious traffic, clogging its tail circuit and interfering with normal traffic. At present, the network operator of a site under attack has no other resolution but to respond manually by inserting filters in the appropriate edge routers to drop attack traffic. However, as DoS attacks become increasingly sophisticated, manual filter propagation becomes unacceptably slow or even infeasible. In this paper, we present Active Internet Traffic Filtering, a new automatic filter propagation protocol. We argue that this system provides a guaranteed, significant level of protection against DoS attacks in exchange for a reasonable, bounded amount of router resources. We also argue that the proposed system cannot be abused by a malicious node to interfere with normal Internet operation. Finally, we argue that it retains its efficiency in the face of continued Internet growth.

cs.NI↗